Non-Technical SummaryThe Korean government has liberalized its foreign exchange market in early April 1999. The foreign exchange law has been simplified considerably: it transformed from a positive list system―no transactions allowed, apart from certain explicit exceptions―to a negative list system―all transactions allowed, with a few exceptional cases. As a sequel, the second stage of the liberalization is scheduled for January 1, 2001. Both liberalization actions aim at removing unnecessary regulations for transactions related to foreign exchange, either directly or indirectly, as well as at increasing depth and liquidity of the market.

The purpose of this study can be put simply into the following questions: What are the special features in Korea's won/dollar foreign exchange market distinguished from foreign exchange markets of other major currencies? Does foreign exchange liberalization increase the trading volume and enhance market efficiency of the foreign exchange market? We believe that the answers for these questions may be able to provide useful policy guide for the relevant parties.
Participants in the won/dollar foreign exchange market (interbank market) are composed of the central bank, authorized foreign exchange banks (dealers) and two foreign exchange brokers. As of November 2000, there are 73 foreign exchange banks, consisting of 21 domestic banks, 50 foreign bank branches and two development banks. Any financial institution that wants to engage itself in foreign exchange transactions should get permission from the government. They should meet minimum requirements of capital, manpower, and facilities.

Currently, two commercial foreign exchange brokers are competing in interbank transactions, the Korea Financial Telecommunications & Clearing Institute (KFTC) and Korea Money Broker Corporation (KMBC). The commercial foreign exchange broker system was introduced to Korea in January 1999. KMBC, a private organization, was allowed to establish a brokerage firm; KFTC, the public foreign exchange broker that enjoyed a monopolistic position in interbank trading, has became a commercial company. These two brokers play an important role in the market. Of all interbank transactions, 96.9% of spot exchange, 63.8% of forward exchange and 93.7% of swap exchange are conducted through both brokers. These numbers are very different from those of other major currencies of which less than 50% of foreign exchange transactions are executed via brokers. Why is brokered interdealer trading a dominant feature in Korea? Simply put, it could be due to the very small size of the foreign exchange market in Korea. Dealers, not wanting to transact directly with other dealers, avoid revealing any trading-related information since the exchange market amounts to only $3 billion or over a day and less than ten banks handle the share of the order flow.

We investigate whether the liberalization measures have induced the market into a more efficient way based on spirits of two very different existing theories: market microstructure theory and market efficiency hypothesis. To test the first one, we use empirical relations between the exchange rate and the trading volume. We see how the market disseminates and reacts to various shocks including private information shocks before and after the liberalization. If the foreign exchange market becomes more efficient after the liberalization, private information shocks should be short-lived or non-existent. We construct the private information shocks as where the exchange rate and the trading volume move at higher than normal revels, respectively. We implement and interpret the shocks within the framework of noisy trading model by Blume, Easley, and O'Hara (1994). The model postulates that information is diffused and incorporated into exchange rate through the trading of informed investors. The uninformed traders infer a new piece of information via the trading volume.
Therefore, the trading process diffuses information as the new piece of information arrives, resulting in a price movement on higher than normal volume. We also take look into how volume shocks and price change shocks affect both the conditional mean and volatility of a variable before and after the liberalization.

We apply the semi-nonparametric (SNP) nonlinear impulse response analysis proposed by Gallant, Rossi and Tauchen (1993) to see how various shocks including the private information shocks affect the variables. The reason we choose the SNP as an empirical tool is that the SNP family of conditional densities is large enough to encompass almost any conditional density, which will minimize the possibility of wrong interpretation of the liberalization measures on account of a specification error. The SNP technique uses Hermite polynomial expansion to directly approximate conditional density. The leading term of the expansion is an ARCH/GARCH. The higher-order terms in the expansion have coefficients which are functions of the conditioning data. In this manner, the polynomial expansion allows for shape deviations from normality and conditional heterogeneity of unknown form. Nonlinear impulse response functions, summarized in Gallant, Rossi and Tauchen (1994), are the extension of the impulse response function of linear VAR to the nonlinear case. In the nonlinear model, the dynamic properties can be elicited by perturbing the vector of conditioning arguments in the conditional density.

Our empirical findings based on nonlinear impulse response functions are as follows. First, large price and volume movements of private information generate persistent responses before and after the liberalization even if the degree of persistence after the liberalization is much smaller than the degree before the liberalization. The results lead us to conclude that the market becomes relatively more efficient after the liberalization, under the framework often going by the name of noisy trading model. That is to say, informed traders may earn much more at the expense of less informed or noisy traders, but the first can enjoy high profits for relatively short duration after the liberalization. Second, the public information shocks do not affect the volatility of the exchange rate in the short run, but the effects do not dampen for a long term in both periods. The same is true for volatility, even though the result is barely statistically significant. Third, the volume shocks or disparate belief shocks are major sources of price deviation from the average exchange rate and persistence in the volatility before and after the liberalization. Therefore, the trading volume itself generates 'excessive volatility' in the market.

We also investigate how the foreign exchange market has changed after the liberalization through the relationship between the offshore NDF (Non-Deliverable Forward) exchange rate and the domestic spot exchange rate. First, we examine the changes in ex-post profit of NDF rates during the sample period. We find the ex-post profit of NDF rate to be very small before the crisis and negative during the crisis period. However, recently the ex-post profit is fluctuating, showing both positive and negative values. It is believed that this comes from the liberalization measure such that domestic banks were allowed to participate in the offshore market since April 1999. Second, we show the status of market liquidity between the three periods―before, during and after the crisis―and find that the liberalization has increased the market liquidity. To examine the liquidity condition, the difference between bid-price and ask-price is analyzed, since there is no available data on the trade volume of the NDF transaction. Due to the liberalization, the liquidity in the forward exchange market is expected to increase as more participants enter. The liquidity increased as the bid/ask difference has reduced about four times between period 1 and period 3. We find that the forward premium puzzle exists in the domestic foreign exchange market. We also find that the influence of crisis still exists in the foreign exchange market. We also obtain empirical results that the offshore NDF exchange rate is not an unbiased indicator. However, the existence of time-varying risk premium explains the bias in the foreign exchange market of Korea. To find the existence of time-varying risk premium, the ARCH-in-mean model is used. ARCH models capture some important aspects of the risk premium in a foreign exchange market. First, the ARCH model is convenient specification for heteroskedasticity, which is an empirical characteristic of exchange rates. Second, most exchange rate changes exhibit leptokurtosis, and the conditional distribution of the Maximum Likelihood procedure of the ARCH model also shows fat-tailed behavior. Lastly, the ARCH-in-mean model is a suitable econometrical model, such that the time-varying risk premium enters into the regression for the conditional mean.

In addition, forward exchange bias clearly appears in the domestic foreign exchange market, indicated by the fact that it did not have a forward premium bias as seen in the main exchange rates in the free-floating exchange rate system. Thus, while most forward exchange rates are known to have a downward bias, an upward bias is observed domestically. This upward bias is thought to be caused by the system collapse following the shock of the financial crisis. Moreover, the upward bias shows that in Korea the interest rate equilibrium condition is satisfied at least in direction. To make an analogy, the government's interest rate stabilization efforts after the crisis greatly helped the exchange rate's downward stabilization. Bias in the domestic foreign exchange market can be explained by checking for existence of a risk premium or peso problem. This means that the domestic foreign exchange market can be understood as the risk premiums of forward exchange buyers and sellers or the expected excess profits (or losses) from expectations of a regime shift.
Based on our empirical finding, we suggest policy implications for exchange and monetary policy. First, the possibility of excessive volatility caused by noise or bandwagon holds a policy implication for the exchange rate system in Korea. Our results show that the degree of excessive volatility in the market grows weaker after the liberalization or adoption of a flexible exchange rate system. It seems to be very contradictory to the general consensus that the exchange rate would be more volatile under a flexible exchange rate system than a fixed one. The clues on the seeming contradiction can be reconciled with the fact that the absolute magnitude of the estimated conditional volatility decreases by more than ten times. As the volatility of the exchange rate decreases, the risk premium become smaller and the incentive for noise traders to participate in foreign exchange transactions shrinks. If the government's policy objective lies in minimizing the exchange rate volatility, one possible candidate for exchange rate system would be multilateral arrangements in the sense of discouraging noisy traders from participating in the market.

Second, indirect intervention of the traditional interest rate policy may not be as much effective as the traditional theory expected. This is due to the risk premium in the determination of exchange rates. The traditional interest rate policy on exchange rate states that increasing domestic interest rates induce more capital inflows and in turn the domestic exchange rates appreciate. However, this would happen only if interest rates and foreign exchange risk premiums were orthogonal. In other words, the exchange rate depreciates rather than appreciates in the case that the increasing interest rate influences the risk premium as a sign of weak economic conditions. According to empirical analysis, the changes of exchange rates depend on the interest rate differential as well as the risk premium. Since the forward rate and realized expected spot rate are correlated positively, as the interest rate differential increases, capital outflows occur and the domestic currency depreciates. As the time-varying risk premium is deemed as an important factor in the foreign exchange market, it is better to understand the effect of risk components in determination of the exchange markets.

Development of the domestic bond market is also another important economic objective for the Korean economy. An efficient and active bond market provides not only effective benchmark rates in the foreign exchange transactions, but also alternative financial assets to diversify the risk. This results in a deeper and more efficient foreign exchange market. In conclusion, the development of foreign exchange market requires a more efficient domestic financial market in general.

Preface

Acknowlegements

Non-Technical Summary

I. Introduction

II. Foreign Exchange Market in Korea
1. Structure and Participants
2. Exchange Rate Regime and Liberalization Measures in Korea
3. Some Stylized Facts of Intraday Volatilities

III. Trading Volume and Exchange Rate Dynamics
1. Trading Volume and Exchange Rate Dynamics
2. SNP Estimation of the Conditional Density
3. Impulse Response Analysis of Nonlinear Models
4. Data and Estimation Results

IV. Testing Efficiency by NDF Exchange Rates
1. Brief Description of the NDF Market
2. Theoretical Background
3. Ex Post Profit of NDF Rates
4. Liquidity
5. Foreign Exchage Market Liberalization and Market Efficiency

V. Summary and Policy Implications

References